@inproceedings{huang-satoh-2026-temporal,
title = "Temporal Tree of Thought: Reasoning-Guided Visual Cue Search for Long-Video Understanding",
author = "Huang, Ziling and
Satoh, Shin{'}ichi",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-emnlp.630/",
doi = "10.18653/2026.findings-emnlp.630",
pages = "13233--13251",
ISBN = "979-8-89176-462-0",
abstract = "Long-video understanding remains challenging for Multimodal Large Language Models (MLLMs) due to limited context length. Uniform sampling may miss crucial moments, while agent-based frame video understanding methods often evaluate frames independently, overlooking the temporal organization of videos. Ideally, evidence selection should mimic how humans answer questions about long videos: first locating the relevant segment from the global context, then zooming into local objects, and details. We propose Temporal Tree of Thought ($T^3$), a training-free framework for adaptive coarse-to-fine long-video understanding. $T^3$ constructs a question-agnostic hierarchical temporal tree via recursive temporally constrained clustering, where each node represents a contiguous segment with an informative key frame. During inference, $T^3$ performs an answer{--}retrieve{--}explore loop: it reasons over coarse representative frames, generates a search statement when evidence is insufficient, and expands relevant branches for finer-grained evidence. This process adaptively shifts the search target from temporal regions to specific objects, and visual details to help video understanding. Experiments on VideoMME, LongVideoBench, and LVBench show that $T^3$ improves Qwen2.5-VL-7B by 0.5{\%}, 4.6{\%}, and 4.4{\%}, respectively, under the same frame budget, demonstrating the effectiveness of structured temporal reasoning. We will release code."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="huang-satoh-2026-temporal">
<titleInfo>
<title>Temporal Tree of Thought: Reasoning-Guided Visual Cue Search for Long-Video Understanding</title>
</titleInfo>
<name type="personal">
<namePart type="given">Ziling</namePart>
<namePart type="family">Huang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shin’ichi</namePart>
<namePart type="family">Satoh</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: EMNLP 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-462-0</identifier>
</relatedItem>
<abstract>Long-video understanding remains challenging for Multimodal Large Language Models (MLLMs) due to limited context length. Uniform sampling may miss crucial moments, while agent-based frame video understanding methods often evaluate frames independently, overlooking the temporal organization of videos. Ideally, evidence selection should mimic how humans answer questions about long videos: first locating the relevant segment from the global context, then zooming into local objects, and details. We propose Temporal Tree of Thought (T³), a training-free framework for adaptive coarse-to-fine long-video understanding. T³ constructs a question-agnostic hierarchical temporal tree via recursive temporally constrained clustering, where each node represents a contiguous segment with an informative key frame. During inference, T³ performs an answer–retrieve–explore loop: it reasons over coarse representative frames, generates a search statement when evidence is insufficient, and expands relevant branches for finer-grained evidence. This process adaptively shifts the search target from temporal regions to specific objects, and visual details to help video understanding. Experiments on VideoMME, LongVideoBench, and LVBench show that T³ improves Qwen2.5-VL-7B by 0.5%, 4.6%, and 4.4%, respectively, under the same frame budget, demonstrating the effectiveness of structured temporal reasoning. We will release code.</abstract>
<identifier type="citekey">huang-satoh-2026-temporal</identifier>
<identifier type="doi">10.18653/2026.findings-emnlp.630</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.630/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>13233</start>
<end>13251</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Temporal Tree of Thought: Reasoning-Guided Visual Cue Search for Long-Video Understanding
%A Huang, Ziling
%A Satoh, Shin’ichi
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F huang-satoh-2026-temporal
%X Long-video understanding remains challenging for Multimodal Large Language Models (MLLMs) due to limited context length. Uniform sampling may miss crucial moments, while agent-based frame video understanding methods often evaluate frames independently, overlooking the temporal organization of videos. Ideally, evidence selection should mimic how humans answer questions about long videos: first locating the relevant segment from the global context, then zooming into local objects, and details. We propose Temporal Tree of Thought (T³), a training-free framework for adaptive coarse-to-fine long-video understanding. T³ constructs a question-agnostic hierarchical temporal tree via recursive temporally constrained clustering, where each node represents a contiguous segment with an informative key frame. During inference, T³ performs an answer–retrieve–explore loop: it reasons over coarse representative frames, generates a search statement when evidence is insufficient, and expands relevant branches for finer-grained evidence. This process adaptively shifts the search target from temporal regions to specific objects, and visual details to help video understanding. Experiments on VideoMME, LongVideoBench, and LVBench show that T³ improves Qwen2.5-VL-7B by 0.5%, 4.6%, and 4.4%, respectively, under the same frame budget, demonstrating the effectiveness of structured temporal reasoning. We will release code.
%R 10.18653/2026.findings-emnlp.630
%U https://aclanthology.org/2026.findings-emnlp.630/
%U https://doi.org/10.18653/2026.findings-emnlp.630
%P 13233-13251
Markdown (Informal)
[Temporal Tree of Thought: Reasoning-Guided Visual Cue Search for Long-Video Understanding](https://aclanthology.org/2026.findings-emnlp.630/) (Huang & Satoh, Findings 2026)
ACL